home | media info | archive | supplier guide | registration | jobfinder | events | about us | contact
OUR WEB SITES
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
March/April 2005
Evaluation: Vocollect Talkman T2: Speech recognition system from Vocollect
Our impression For this issue of m.logistics, we decided to devote our Evaluation feature to finding out why everyone seems to be talking about voice recognition technology, and how well it actually works. We don't have a warehouse to manage, of course, but we did try out the Vocollect Talkman T2 SR-20 headset linked to a warehouse management system, and we witnessed a demonstration of a picking operation that compared voice, paper and scanning picking operations. This was arranged by Psion Teklogix, which sells Talkman T2 systems in its supply chain solutions packages, and is currently particularly keen to promote voice-related technology. The Talkman 2 is a dedicated voice terminal consisting of an SR-20 or SR-30 headset (SR for stands for Speech Recognition) linked by a thin cable to a terminal worn in a waist holster. The terminal uses speech recognition and speech synthesis technologies to provide real-time communication with the WM or ERP system. The internal architecture supports 802.11b Wi-Fi compliant radios. 
A lot of thought has gone into the ergonomics of the headset design, and it shows. The headband has a double strap for extra adjustability and weight distribution, and there is ample padding, so it can be worn comfortably for a full length of a shift. It can be worn on either ear, and doesn't interfere with spectacles, caps or cold-weather gear. The headset incorporates a three-position "Speech Groove" memory that enables users to adjust it repeatedly to their preferred wearing position. The cable linking the headset to the terminal has what Vocollect calls soft connectors - which means that if the wire gets caught or trapped, the connectors snap off. This reduces risk of injury and also protects the equipment from damage. The voice recognition engine was developed by Vocollect and is speaker-dependent - which means you must train it to recognise your voice. This takes between 10 and 20 minutes depending on the amount of vocabulary required (most pick operations require between 10 and 40 words and numbers). It is single-word based, so you only have to say "ready," for example to tell Talkman you are ready to start picking. You don't have to add wake-up or close-down words at the beginning and end of each operation. It's all very straightforward, and certainly no harder than trying to remember which buttons on the barcode scanner to press. Because the terminal is trained to recognise individual voices it won't confuse your voice with that of a colleague standing close by. If you want a chat with a co-worker, you can also tell Talkman to go to sleep. You can slow down the text-to-speech engine speed for trainees. Employees who aren't sure if they've understood a spoken instruction can also ask the system for a description of the product to be picked, and can check any location or item number. This feature helps productivity by eliminating mis-picks, and also helps make Talkman 2 more user-friendly. You do not feel as if a disembodied, uninterested voice is controlling your working life. For Psion Teklogix's demonstration, identical aisles were set up, each with 80 cases of goods. One worker had the voice-picking system, another used a conventional barcode-based picking system, while the third used paper-based picking. Each had to pick 20 items. The demonstration showed how impressively fast voice-directed picking can be. The voice-pick task was completed in just over 5.07 minutes, barcode scanning took 6.5 minutes, and the paper pick took 8.22 minutes, not including the time the warehouse supervisor has to spend manually entering any details into the WMS. Speech is sometimes seen as an impediment to productivity. Managers are worried that employees will become irritated by a disembodied voice talking constantly to them. In reality, though, speech comes in short bursts, either from the system as it directs the picker to the next location, or from employees, confirming check digits and the number of items picked, for example. Psion says that on an eight-hour shift, 30 minutes of speech is typical. If you're worried about the possibility of technophobe resistance to VR technology, reports are that worker reaction is exactly the opposite of this; they actually like the idea of using comparatively hi-tech equipment in a staunchly traditional environment. Watching the demonstration, we also found it obvious that with voice, the picking task was actually easier and less stressful physically - advantages that are bound to go down well with employees. Our verdict The headset is light and easy to wear, and the adjustable headband and microphone are both big pluses in the comfort stakes. Although our experience of the text-to-speech function was limited to the Psion Teklogix demonstration, we can confirm that the synthesised voice is clear and distinct. You really do get the impression you're interacting with the system as you would with a keyboard or handheld computer. The demonstration showed that claims of improved productivity really can be justified on multi-pick operations. More than that, it was also very clear that using voice actually makes the picking job a lot easier because workers can focus on where they're going, and have both hands free to collect the items on the list.
|